Buckminster
Well-Known Member
$50k sounds about right although I would rather Tesla leases them out for $25k/annum. Bots could spend 2 years in a home, 5 years as a cleaner at a school or shopping mall and then another 10 in a factory.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Tesla Optimus Sub-Prime Robot
- Thread starter MC3OZ
- Start date
Yeah, I thought they would lease them out at about $10/hour. Which works out to about $20K/yr if they are only used 8 hours a day, five days a week.
Just $25K/annum is actually low balling them. These things will first be used in higher value work, like factory work, something that costs factory owners $100K/yr per person in fully loaded costs.
So they are worth about $500K in actual value (if they can replace a human is whatever task they are given).
Good video comparing optimus' competition:
My analysis:
Elon is now in charge of optimus. Elon's past projects generally follow the same pattern. Quickly get a minumum viable product to the market. Improve cost and functionality by scale. Get real feedback and quickly iterate. Focus the development on things that are useful long term and that increases the speed of development. Introduce costs functions/score and let the team do whatever they want that optimises the score, improve the cost function/score to catch things they realized they missed whenever they find problems. Automate as much as possible and remove things that are not needed aggressively. Have a war time mentality, push deadlines and be extremely optimistic then force the team to work hard and use all their talent to get there.
The other teams do less of this. They want to do things they can show off to impress their collegues, are academic or things that are a fun intellectual exercise. Rather than cost functions they use intuition and best practices. Products are figured out later, primarily it's about doing the research. Feedback is not needed and real customers are messy. Slow and steady wins the race.
My analysis:
Elon is now in charge of optimus. Elon's past projects generally follow the same pattern. Quickly get a minumum viable product to the market. Improve cost and functionality by scale. Get real feedback and quickly iterate. Focus the development on things that are useful long term and that increases the speed of development. Introduce costs functions/score and let the team do whatever they want that optimises the score, improve the cost function/score to catch things they realized they missed whenever they find problems. Automate as much as possible and remove things that are not needed aggressively. Have a war time mentality, push deadlines and be extremely optimistic then force the team to work hard and use all their talent to get there.
The other teams do less of this. They want to do things they can show off to impress their collegues, are academic or things that are a fun intellectual exercise. Rather than cost functions they use intuition and best practices. Products are figured out later, primarily it's about doing the research. Feedback is not needed and real customers are messy. Slow and steady wins the race.
Last edited:
That was a great video. Worth the time if you want to get up to speed on the entire humanoid robotic market. A couple of interesting things I learned:
Boston Robotics‘ Atlas. I knew it was going to have crappy specs becasue the demos are so flashy, but I was still surprised how crappy. Only 4’11 high, yet weighs 200 pounds, only 20 degrees of freedom, with a battery runtime of just one hour. Not designed for manufacturing, it’s only a prototype, and uses minimal AI.
Agility Robotics’ Digit is the most impressive IMHO. At 5’9” and 140 pounds, 35 pounds payload, 1.5 to 3 hours runtime, fast charging (16 hours/24 runtime), built for manufacturability, furthest along (will soon be in customer hands).
There are others that are doing really interesting things. Most are very well capitalized. OpenAI seems to have a great hand coupled to a great AI. Teslabot has great specs, but they are still early and may run into issues that take time to work out (like it has for FSD).
It is very much early days in this space and there will likely be several bots entering into the workforce, each specialized in their particular area (kinda like different types of cars/trucks for different purposes). It’s going to be an interesting ten years!
No one is so far ahead that they’ll blow anyone
Happened upon this article awhile ago, maybe its helpful?
Robotics Outlook 2030: How Intelligence and Mobility Will Shape the Future
Established robotics companies must counter startups that are innovating in areas with the potential to generate high profits and alter the dimensions of the industry.
www.bcg.com
A funny commentary on the state AI is in right now…
jeewee3000
Active Member
I think Wiferion tech makes more sense for EV's.
Optimus should be capable enough of plugging a charger into itself. (If it can't handle that task, it's pretty much useless for most other tasks)
And cable will always be more efficient/potent than wireless chargers.
Unless you install wireless chargers in most places where certain Optimi operate, and they stay more or less charged during work. (for example, for routine tasks Optimus has to walk between point A and point B, and it remains at point A for a minute every time, then you could have Optimus charge wirelessly at point A to lengthen operation time greatly. But again, this is only useful if the amount of time for Optimus to plug in/unplug is a substantial amount of the amount of time spent at point A. If Optimus stays in location for let's say 30 minutes at a time, it's worth it to let Optimus plug in first and stay plugged in during operation at point A.
EV's on the other hand -> wireless charging embedded in parking lots is a dream come true. (destination charging)
RoboCat: A self-improving robotic agent
Robots are quickly becoming part of our everyday lives, but they’re often only programmed to perform specific tasks well. While harnessing recent advances in AI could lead to robots that could...
Buckminster
Well-Known Member
Wow, this is awesome. I told my wife and she said "Let me know when it can pull weeds"
Buckminster
Well-Known Member
Summary of the video:
00:00:00 - 00:45:00
In this YouTube video titled "LEAK! Tesla Optimus BOT Revelations Show Production Only Months Away; Scott Walter Reports Huge News," the host and guest discuss the progress and potential of the Tesla Bot over the past seven months. They reflect on the initial skepticism surrounding the bot, but with the emergence of working prototypes and demonstrations of coordination, it has gained more credibility. The possibility of using bots in Tesla's factory for automation is also discussed, although it remains unclear to what extent they are currently involved in production. Speculation is made about the number of bots being used, with estimates ranging from dozens to hundreds. The video also mentions the importance of software and AI in achieving automation, with Tesla's Dojo technology setting them apart. The discussion touches on the responsibilities of a manufacturing line position at Tesla, highlighting the tasks involved in operational efficiency and equipment integration. The future tasks that the Tesla Bot may undertake during production, such as replacing empty totes and potentially signaling humans for assistance, are mentioned. The potential for the bot to perform gymnastic-type movements is also discussed, suggesting Tesla's pursuit of this capability due to their competitive nature and talented robotics team. Challenges in automating the process of connecting the wiring harness in cars are addressed, with Tesla developing a bot to work with flexible wiring. The impact of Tesla Bots on the company's bottom line in terms of cost savings and improved efficiency is emphasized. The potential savings and impact on margins resulting from the introduction of Tesla Bots are discussed, with the speaker suggesting that Wall Street should consider Tesla beyond just a car company. The disruptive nature of electric vehicles, particularly the Tesla Bot, is compared to the transition from a farming to an industrialized economy. Overall, there is excitement about the innovation and its potential to revolutionize transportation and create wealth and efficiency.- 00:00:00 In this section, the host Randy Kirk and guest Scott Walter discuss the history of the Tesla Bot and its progress over the past seven months. They reflect on how the initial skepticism surrounding the Tesla Bot has gradually shifted as more evidence emerged, such as the working prototypes seen during Investor Day and the coordination of multiple bots during Shareholder Day. They also mention the speculation regarding the bots' usefulness and mention the low-hanging fruit tasks, like loading simple parts, which the bots are expected to perform. Overall, they highlight the excitement of being part of the history and development of such a groundbreaking invention.
- 00:05:00 In this section, the speaker discusses the use of Bots in the factory and the potential for automation. While it is not clear if the Bots are currently involved in assembling cars or just demonstrating tasks, it is revealed that there are already Bots operating in the factory. The speaker speculates that these Bots could potentially replace employees in the future, but it is uncertain how many Bots are currently being used. The importance of software and AI in achieving automation is emphasized, and it is noted that Tesla's Dojo technology sets them apart from others in terms of vision and language processing. Overall, the existence of Bots in the factory sparks excitement and speculation about the future of automation at Tesla.
- 00:10:00 In this section, the discussion revolves around the number of robots Tesla may be using in their production process, with estimates ranging from dozens to hundreds. The cost of building these robots is considered minimal, so the limitations on their numbers may be due to factors such as testing readiness or the difficulty of assembling certain components. The location of the manufacturing facility is speculated to be in Palo Alto, where job openings for Tesla bots have been posted. However, the hiring of a Sparks position indicates that the production line is still being set up, which could take anywhere from nine months to a year. Overall, it is predicted that Tesla may be ready to start producing the Tesla bots by the middle of next year.
- 00:15:00 In this section, the speaker discusses the job opening at Tesla for a manufacturing line and highlights the tasks and responsibilities involved. They mention the need for material flow and operational efficiencies, equipment integration, management of suppliers, and the development of manufacturing processes. The speaker speculates that hiring someone with experience in automation equipment could accelerate the production timeline, estimating that it may take around nine months to a year for full production readiness. They also suggest the possibility of a few hundred robots being produced by the end of the current year and potentially thousands by the end of next year. The speaker then introduces the concept of Bots already performing simple tasks in the Fremont factory, such as picking and placing, with the aim of being able to recognize when parts need replenishing.
- 00:20:00 In this section, the speaker discusses the potential tasks that the Tesla Optimus BOT may be involved in during production. One task mentioned is the process of identifying and replacing empty totes with full ones to ensure a continuous supply of parts. The speaker speculates that the bot may be responsible for signaling when a tote is empty and instructing a human worker to replace it with a full one. This task could potentially slow down the overall production cycle. The speaker also mentions a debate that took place about whether the bot should be taught gymnastic-type movements or more dexterity-based tasks. Despite some skepticism about the bot's current capabilities, the speaker suggests that it may have more potential than meets the eye, especially considering its powerful motors.
- 00:25:00 In this section, the speaker discusses the possibility of Tesla's Optimus robot being able to do backflips or other gymnastic movements. They speculate that Tesla's competitive nature and the talented robotics team working at the company may be motivating factors for pursuing this capability. The speaker also mentions the importance of improving the robot's walking mechanics, specifically the toe-off motion, in order to perform such maneuvers effectively. Additionally, they touch on the genesis of the idea for Optimus, suggesting that Elon Musk may have had the concept in mind for a long time and decided to pursue it during the Model 3 ramp-up.
- 00:30:00 In this section, the speaker discusses the challenges of automating the process of connecting the wiring harness in a car. They mention that the wiring harness is a manual operation and it has been difficult to automate due to its flexibility and the need for precision in connecting the wires. However, they reveal that Tesla is working on developing a bot that can work with flexible wiring and that they seem to be further along in this process than what has been shown in previous videos. The speaker also emphasizes the importance of a good vision system for this task and suggests that if Tesla can develop a real-time vision system that can adaptively control the process, they can achieve a flexible and fixtureless process. The speaker, who has experience in robotics, discusses how hiring a lead position for a production line usually indicates that the company is planning to begin production within the next 12 to 18 months.
- 00:35:00 In this section, the speaker discusses the plans and progress of Tesla's Optimus BOT project, stating that they are actively looking for someone to implement their vision and start hiring a team. They also mention that the project is likely to be based in Sparks, rather than Palo Alto. The speaker addresses concerns about the impact of Tesla Bots on the company's bottom line, noting that the focus should be on cost savings rather than direct sales. They explain how Tesla Bots can significantly reduce costs in operations and highlight the potential value-add of automating certain tasks. The speaker provides an example of how much a worker's operation can cost and emphasizes the cost savings associated with Tesla Bots.
- 00:40:00 In this section, the speaker discusses the potential savings and impact on margins that could result from the introduction of Tesla Bots in Tesla's production process. By eliminating workers and increasing automation, Tesla could save a significant amount of money, which would directly contribute to their bottom line. The speaker suggests that Wall Street should take this into consideration when evaluating Tesla as more than just a car company. The conversation also touches on the contentious nature of automation and its potential to eliminate jobs, as seen in past examples like email and word processing. However, the speaker argues that the benefits of automation, such as cost savings and efficiency, outweigh the potential job losses.
- 00:45:00 In this section of the video, the speaker discusses the transition from a farming economy to an industrialized one, and how that led to the creation of new industries and more plentiful resources. They argue that the same disruptive process is happening now with the rise of electric vehicles, particularly Tesla's Optimus BOT. The speaker suggests that this innovation will revolutionize transportation and break the economy in a positive way by creating more wealth and efficiency. They emphasize the importance of this invention, comparing it to the wheel and fire, and express excitement about being part of this massive change.
Last edited:
I think I figured out how a lot of recent developments connect. We have Ashok showing foundational models where the vision system sees the world and understands it.
You can prompt it for actions and it will predict the future. Ok, cool, but why? To make the neural network a little bit more efficient? Sure that's nice but is that all? No I think there is a bigger picture. On one of his slides it says:
Think back at investor day where they demoed a human teaching the bot how to perform a task.
The human had a backpack with custom hardware and cameras on his head. Are users gonna have to buy one of those kits to show the robot what to do? What does it actually do? Basically the gear estimates a few poses for some joints and the end effectors(fingers). Maybe there's a better way of getting those poses with having all that extra hardware? How about getting it from the world model?
This is how they will do it:
Have the robot stand next to a human observing the human demonstrating the task. The robots understand the world to understand where the human is placing its fingers and from the context it understands that the human is gripping the objects. It understands what kind of object the human is manipulating, its properties and figures out if it is heavy and it can predict what will happen given different prompts.
Is it hard to learn from example? You can do it, proving that it's possible to learn to do it.
So I think what Tesla will show during AI Day 2023 is a human demoing a list of actions and the robot then performing the same actions by itself. After the observation the robot will build a world model of what just happened. Then it will simulate itself doing the same subtasks, play around with some assumptions if it's not sure about the friction, mass, rotational inertia, bending of cables etc, practice in simulation and find a good efficient strategy and generate the control sequence that it can later adapt as it executes the plan. Practice a bit in simulation, the walk up and perform the task. Not unlike how you would try to solve a task someone just showed you or how a climber is observing a route:
- Useful as a Neural Net Simulator
Think back at investor day where they demoed a human teaching the bot how to perform a task.
This is how they will do it:
Have the robot stand next to a human observing the human demonstrating the task. The robots understand the world to understand where the human is placing its fingers and from the context it understands that the human is gripping the objects. It understands what kind of object the human is manipulating, its properties and figures out if it is heavy and it can predict what will happen given different prompts.
Is it hard to learn from example? You can do it, proving that it's possible to learn to do it.
So I think what Tesla will show during AI Day 2023 is a human demoing a list of actions and the robot then performing the same actions by itself. After the observation the robot will build a world model of what just happened. Then it will simulate itself doing the same subtasks, play around with some assumptions if it's not sure about the friction, mass, rotational inertia, bending of cables etc, practice in simulation and find a good efficient strategy and generate the control sequence that it can later adapt as it executes the plan. Practice a bit in simulation, the walk up and perform the task. Not unlike how you would try to solve a task someone just showed you or how a climber is observing a route:
Buckminster
Well-Known Member
Buckminster
Well-Known Member
Should be doing useful things in Tesla factories in 2024.
Optimus arms and legs can be used to create a Cyborg body combined with Neuralink.
Buckminster
Well-Known Member
RT-2: New model translates vision and language into action
Introducing Robotic Transformer 2 (RT-2), a novel vision-language-action (VLA) model that learns from both web and robotics data, and translates this knowledge into generalised instructions for...
Some examples of commands it can handle:
Here is how it translates instruction to plan(note that they all start with 1 and end with 127)
Similar threads
